Banking Can’t Afford to Wait on AI
Most of us expect money to go where we tell it to. The common mental model of monetary transactions is that money moves at the speed of light. But everything in the world is some kind of engineering, and we rarely think about the systems working behind the scenes.
Yet, having spent much of my career in financial infrastructure, I know that what appears to be a simple transaction can involve a surprisingly complicated network of applications, payment infrastructure, and legacy technology. A 2025 U.S. government report found that 8 out of 11 critical legacy IT systems examined still used outdated programming languages.
One of the biggest mistakes I see organizations make is looking at modernization as a piecemeal affair. Companies replace one application or install new software and assume they are done for the next decade.
Installing a modern system does not necessarily make the whole infrastructure modern. I have seen how ancient technology like file-based integrations and CSV files is used to connect systems, and, in my experience, these dependencies can and often do produce legacy issues while modernization takes place.
In core banking, these issues come to the fore immediately. I have seen the scale of these undertakings, including the work I did back in 2016 at Brazil’s Bradesco to rebuild its core banking account-management platform. Experiences like this reinforce my belief that banks cannot rely solely on large-scale replacement projects. Instead, they need to manage their entire technology footprint as a portfolio, continuously prioritizing upgrades, improvements, and replacements to deliver the greatest long-term value. This becomes particularly important as new forms of money emerge.
Consider a customer transferring money from a deposit account to an investment product offered by the same banking group. Even when both businesses belong to the same group, they may operate as separate legal entities, requiring funds to travel over an external payment rail before returning to the bank. Tokenized deposits could enable eligible transactions to happen within the bank’s own infrastructure.
I believe tokenized deposits could have a significant practical impact within banks over the next five to ten years. Stablecoins could extend some of these capabilities beyond an individual institution. Once funds are tokenized, converting them into a stablecoin could facilitate transfers between banks and support business-to-business payments.
However, the challenge I find most pressing is the emergence of AI agents capable of initiating and executing financial transactions. We have moved rapidly from using AI to generate documents and summarize information to connecting it with email, calendars, and purchasing activities. Increasingly, agents can interact with websites, compare prices, and transact with other agents.
I believe the next logical question facing the banking industry is to ask whether payment systems that were originally designed around humans are ready for potentially thousands or millions of automated agents transacting across different interfaces. When a transaction passes through several agents, how do we establish who initiated it, who authorized it, and who ultimately bears responsibility?
Without reliable traceability, there is a fraud risk. Someone could authorize an agent to make a purchase and later claim they had no control over its actions, potentially seeking a reversal after receiving the goods. At scale, these situations could become costly and difficult to resolve. In my view, authorization, auditability, and reporting cannot be afterthoughts. We need to build them into the transaction process itself.
But there is a way to resolve this. Imagine businesses assigning AI agents definite permissions within the network or a blockchain. Imagine banks automating suitable internal transactions while enabling faster transfers between institutions. These efficiencies could benefit businesses and customers alike. But they depend on systems that can process transactions reliably and establish a clear chain of authorization.
That is why I believe banks, payment networks, standards organizations, and merchants must act now. Banks need to modernize continuously rather than wait for their next major technology project. Standards bodies and industry groups need to accelerate practical frameworks for agentic payments. Merchants must make their online systems accessible to AI agents while ensuring transactions remain secure and accountable.
Technology is advancing faster than standards can keep pace. I believe the future of finance depends on building infrastructure and governance that enable secure, traceable, and accountable transactions while keeping innovation moving forward.
About the Author
Craig Borysowich is a financial infrastructure professional with experience in banking technology, payment systems, and large-scale infrastructure modernization. Through his work with IM Edge, he focuses on helping organizations improve the delivery of infrastructure transformation and adopt a more continuous approach to technology modernization.
